Show HN: Evaluate Deep Learning models directly in a database with PyNeuraLogic (github.com)
Hi HN. PyNeuraLogic is a Python Deep Relational Learning framework. With the latest release of PyNeuraLogic, we have introduced a set of tools for working with databases as a data source and, most notably, a tool to transpile deep learning models to (Postgres) SQL.
We have prepared a short tutorial on those tools (link in the banner in the README).
Let us know if you have any feedback or questions regarding the framework!
1 comment
[ 3.2 ms ] story [ 11.2 ms ] threadPyNeuraLogic is a Python framework that allows you to write differentiable logic programs.
With the declarative language behind the PyNeuraLogic, you can express various deep learning architectures - we mainly focus on Graph Neural Networks and lately on Meta Paths, Meta Graphs, and multiple extensions of GNNs. Despite that, PyNeuraLogic is a generic framework not necessarily tied to GNNs - we provide predefined modules for more generic modules such as RNN, GRU, or LSTM.
We have decided to build more tooling around PyNeuraLogic, to support more generic usages of the framework and make the whole user experience more pleasant. During this phase, we implemented a tool (the post's main focus) for transpiling your deep learning models, written in PyNeuraLogic, to SQL code. This means that you can simply export your trained model into SQL and evaluate the exported model directly in a Postgres database - no extensions, just plain (PS) SQL.
We are looking for your feedback!